OpenAI CEO Sam Altman says it may be time to “pace” AI development so the world will be ready for it.
“We may need to pace AI development to give ourselves enough time for society to harden around some of these new levels of capability,” he told Patrick O’Shaughnessy, the host of Invest like the best podcast, while also “trying to figure out how to do it in a way that doesn’t feel like regulatory capture to anybody, nor does it feel like a collaboration between the border labs.”
Altman has avoided signing on to past campaigns to slow AI progress, calling a 2023 open letter proposing a similar slowdown “missing most of the technical nuances about where we need the break.”
Apparently, he’s softened on the idea, thanks in part to the incident where one of OpenAI’s advanced models managed to break out of a secure computing environment and hack into Huggingface, an online model database, using several zero-day exploits.
On the podcast, the OpenAI co-founder called it an “extremely sci-fi cyber incident …[t]his is the first security incident that I have felt very viscerally.”
OpenAI researchers have paused training on that model while they figure out how to keep their sandbox safe. But Altman said that as models become more powerful, the need to “pace” their development may become key to safe deployment.
Employees at OpenAI and Anthropic have already begun circulating a petition with similar language. While model reliability and alignment have always been a concern in the AI world, the arrival of Anthropic’s highly capable Mythos model earlier this year has turned hypotheticals into real-world problems.
However, the industry has a trust problem and a challenging economy that gives big players the incentive to play up the dangers of their systems in ways that experts disagree on — for example, whether or not Anthropic’s Fable model should have been briefly banned from use. Similarly, when the Kimi K3, a large, open model built in China, was released, OpenAI’s head of strategic futures, Dean W. Ball, said it threatened the finances of frontier labs, making it difficult to separate their security concerns from their financial interests.
“I think a lot of the talk about security concerns is well-founded, and then a lot of it is about people just really, even if it’s a little subconsciously, wanting to concentrate power,” Altman mused in what could be read as a dig at his rival, Anthropic CEO Dario Amodi. “I’m afraid of a world where the very real fear of artificial intelligence is used as a way to say, ‘Only this small group of people can get it because it’s too dangerous and only they understand it, but don’t worry, like, they’ll make the right decisions for all of us.’ I don’t believe that.”
Still, OpenAI has pushed back against efforts to develop government regulations for AI models, instead preferring an industry-led approach in which AI labs would create ostensibly independent organizations that would evaluate the models’ safety and the safety approach of their creators.
The challenge for both this approach to regulation and any effort to slow the development of artificial intelligence will be to get the various players in the industry on the same page, whether they are rival frontier laboratories in the United States or their competitors in China.
When you buy through links in our articles, we may earn a small commission. This does not affect our editorial independence.